What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
This subagent crawls the Tracklib “New Arrivals” page and extracts up to 20 song entries containing {title, artist, tracklib_url}.

B) FINAL TASK OUTPUT  
A structured JSON array (or similarly typed data structure) containing up to 20 entries, each entry with the fields:  
{  
  "title": "...",  
  "artist": "...",  
  "tracklib_url": "..."  
}

C) SUBAGENT INPUT  
1. The fixed URL "https://www.tracklib.com/music/new-arrivals" (no variation needed).  
2. Text instruction specifying the exact fields to be extracted, for example:  
   "Extract song title, artist, tracklib_url for up to 20 new arrivals."

E) SUBAGENT TASK SUMMARY  
1. Receive the subagent input, which is the webpage URL plus the extraction instruction.  
2. Pass the URL and the extraction instruction to skill #226 (Extract Structured Data From 1x URL).  
3. Use skill #226 to scrape the page and gather the 20 (or fewer if only fewer are available) most recent songs, returning structured data with {title, artist, tracklib_url}.  
4. The output from skill #226 is then taken as the subagent’s final output: a JSON structure of up to 20 items with the required fields.

F) SILOS  
• SILO #1 – Scrape and Return Structured Data  
   → Input: URL + Text instruction   
   → Action: Use skill #226 to gather and parse the webpage content, extracting up to 20 songs   
   → Output: [latest-track-metadata] (JSON array with up to 20 items)

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent takes the user’s brief text description and a list of up to 20 latest track metadata entries, then filters and selects exactly 3 that best match the user’s description.

B) FINAL TASK OUTPUT  
A JSON list of the top 3 selected tracks, each with {title, artist, tracklib_url}.

C) SUBAGENT INPUT  
1. [variable1]: A short user description (e.g., “Looking for soulful strings and anything that is multitrack”).  
2. [latest-track-metadata]: A list (up to 20 items) of song metadata objects {title, artist, tracklib_url}.

E) SUBAGENT TASK SUMMARY  
1) The subagent receives two inputs: the user’s description ([variable1]) and the scraped track metadata list ([latest-track-metadata]).  
2) It uses Skill #223 (Powerful LLM Prompt-to-Text Response) to process these inputs. The subagent constructs a prompt instructing the LLM to:  
   – Analyze the user’s description.  
   – Review the provided track metadata list.  
   – Select exactly three tracks that best align with the user’s description.  
   – Return them in a short JSON list format, each item having the structure {title, artist, tracklib_url}.  
3) The skill returns the final filtered list (top 3 tracks).  
4) The subagent’s final output is this JSON list.

F) SILOS

• Silo 1: Gather Inputs  
   – Input: [variable1] (user’s description)  
   – Input: [latest-track-metadata] (list of song metadata)

• Silo 2: Filter Tracks  
   – Action: Use Skill #223 (Powerful LLM Prompt-to-Text Response)  
     – Prompt includes: user’s description + the track metadata list, asking for the best 3 matches in JSON.

• Silo 3: Produce Final Output  
   – Output: The top 3 song entries in JSON format, e.g.,  
     [  
       { "title": "...", "artist": "...", "tracklib_url": "..." },  
       { "title": "...", "artist": "...", "tracklib_url": "..." },  
       { "title": "...", "artist": "...", "tracklib_url": "..." }  
     ]

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent takes up to 3 filtered songs (each with title & artist) and, for each one, locates a YouTube link (via a web search) then downloads the audio from that link as an MP3 file.

B) FINAL TASK OUTPUT  
Three MP3 URLs (one per selected track). Example final output data structure: ["https://…mp3", "https://…mp3", "https://…mp3"].

C) SUBAGENT INPUT  
• [filtered-track-list]: Array/list of up to 3 entries, each containing { title, artist, tracklib_url }.

E) SUBAGENT TASK SUMMARY  
1) Parse the [filtered-track-list] to get each track’s title & artist.  
2) For each track (repeat up to 3 times):  
   a) Use SKILL #224 (Oracle Ask A Question) with a query like “Find the best YouTube link for: <title>, <artist>, official audio.” The Oracle will likely return one or more relevant YouTube links.  
   b) Select the top or most relevant YouTube link from the Oracle’s response.  
   c) Pass that YouTube link into SKILL #193 (Download Entire YouTube Video As MP3), which outputs an MP3 URL for that track.  
3) Aggregate the 3 resulting MP3 URLs into a final list.  
4) Output that list as the subagent’s final result.

F) SILOS  
• Silo 1: “Locate YouTube link”:  
   – Action: For each track, run SKILL #224 by prompting “<title> <artist> official audio.”  
   – Outcome: The best YouTube URL for the track.  
• Silo 2: “Download MP3 from YouTube link”:  
   – Action: For each identified YouTube URL, run SKILL #193.  
   – Outcome: MP3 file URL.

The subagent completes when it has three MP3 URLs (one for each selected track), returning them as its final output.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
This subagent takes in up to three MP3 URLs, analyzes each track’s tempo and beats, then cuts a 16‐beat snippet from each track and returns the snippet URLs in WAV format.

B) FINAL TASK OUTPUT  
A structured list (array) of up to three WAV snippet URLs (one per MP3 track), where each snippet is the first 16 beats of the song.

C) SUBAGENT INPUT  
• A list of up to 3 MP3 URLs (e.g. ["https://example.com/audio1.mp3","https://example.com/audio2.mp3","https://example.com/audio3.mp3"])

E) SUBAGENT TASK SUMMARY  
1) Subagent receives up to 3 MP3 file URLs.  
2) For each MP3:  
   (i) Use Skill #180 (Extract Beatpoints & Tempo of MP3) to get the tempo (BPM) and a list of beat offsets.  
   (ii) From these beat offsets, determine the end time of the 16th beat.  
   (iii) Use Skill #219 (Cut Wav/mp3 Audio into Multiple Pieces/Samples) with the MP3 URL as input and the determined time range (start=0, end = offset of 16th beat).  
   (iv) Skill #219 returns a WAV snippet URL.  
3) Collect all 3 WAV snippet URLs into a simple list or JSON array as the final output.

F) SILOS

• SILO #1 (for Track #1):  
  1. Input: MP3 #1 URL  
  2. #180 → Retrieve beat offsets & BPM for MP3 #1  
  3. Determine offset of the 16th beat (e.g., t16)  
  4. #219 → Cut the audio from 0s to t16 → returns 1x WAV snippet URL

• SILO #2 (for Track #2):  
  1. Input: MP3 #2 URL  
  2. #180 → Retrieve beat offsets & BPM for MP3 #2  
  3. Determine offset of the 16th beat  
  4. #219 → Cut from 0s to that offset → returns 1x WAV snippet URL

• SILO #3 (for Track #3):  
  1. Input: MP3 #3 URL  
  2. #180 → Retrieve beat offsets & BPM for MP3 #3  
  3. Determine offset of the 16th beat  
  4. #219 → Cut from 0s to that offset → returns 1x WAV snippet URL

→ Combine all three WAV snippet URLs in a final list/array, which is the subagent’s output.

4 Template & Links

Expand Flow

A) SUBAGENT SUMMARY  
This subagent (#5) consolidates the final results of the entire workflow (the three chosen tracks, their MP3 URLs, and the 16-beat clips) into one single text summary.

B) FINAL TASK OUTPUT  
A single text response (JSON or plain text) providing a cohesive summary of:  
• The three selected tracks (title, artist, tracklib link)  
• MP3 URLs for each full track (from Subagent #3)  
• The 16-beat audio clip URLs for each track (from Subagent #4)

C) SUBAGENT INPUT  
1. [filtered-track-list]: A short text listing of the top 3 chosen songs (title, artist, tracklib_url).  
2. [mp3-download-urls]: The final MP3 URLs for each of the 3 songs.  
3. [16-beat-clip-urls]: The 16-beat audio snippet URLs for each of the 3 songs.

E) SUBAGENT TASK SUMMARY (Detailed Flow)  
1. Subagent receives the three inputs:  
   • [filtered-track-list]  
   • [mp3-download-urls]  
   • [16-beat-clip-urls]

2. Use Skill #223 (Powerful LLM Prompt-to-Text Response) to generate a final summary.  
   • Construct a prompt that includes the three inputs (track info, MP3 URLs, snippet URLs) and requests a concise summary.  
   • The LLM (#223) outputs a single text block that clearly lists the songs, the full-track MP3 links, and the sample snippet URLs.

3. Subagent final output:  
   • [final-summary]: A single text response summarizing all the provided information (the 3 songs chosen, their MP3 URLs, and their sample snippets).

F) SILOS  
This subagent’s workflow is straightforward enough that it only comprises a single silo:  
• Silo 1 → Calls #223 with all inputs, returns [final-summary].

5 Template & Links

Expand Flow

A) SUBAGENT SUMMARY  
This subagent’s role is to take the final clipped audio files (the three 16-beat snippets generated for each of the chosen songs) and package them up—along with any relevant metadata—into a single final deliverable (e.g., a .zip file or a neatly organized final output). Because there is no built-in “zip” skill, this subagent will conceptually handle the “bundling” step, returning a single final artifact and any text summary describing it.

B) FINAL TASK OUTPUT  
The final subagent output will be:  
• A single “packaged” artifact (conceptually a .zip or similarly bundled set) containing the 3×16-beat snippet audio files.  
• A short text summary (or metadata file) explaining the included audio snippets (e.g., which song each snippet is from, BPM, etc.).  
Because there is no direct “zip” skill, the subagent will conceptually produce or reference one final “packaged” location for all these snippet files, plus summary text.

C) SUBAGENT INPUT  
1. The 3 final clipped audio snippet URLs (16-beat samples) from Subagent #4 (the BeatChopperAgent).  
2. The textual metadata needed for each snippet (e.g., selected track’s title, artist, and any tempo data deemed relevant).  
3. (Optional) The user’s original or additional formatting instructions for how the summary or packaging should look.

D) SUBAGENT TASK SUMMARY  
Below is the step-by-step breakdown of the subagent’s internal flow. This subagent minimally needs to produce one final deliverable—an organized distribution of the snippet files plus a short text description:

1. (Receive Inputs)  
   – Inputs arrive:  
     • [16-beat-clip-urls] (the three audio snippet URLs)  
     • [filtered-track-list] or other relevant metadata from previous agents (song title, artist, tracklib_url, BPM, etc.).  
   – No skill invocation is required just to receive these; they are used in subsequent steps.

2. (Generate Short Metadata Summary)  
   – (Optional) If a straightforward textual summary of the final snippet set is required (e.g., listing snippet #1 from “Song A,” snippet #2 from “Song B,” BPM info, etc.), the subagent can call:  
     • SKILL: #223 (Powerful LLM Prompt-to-Text Response)  
       INPUT: A prompt containing the track metadata, snippet URLs, and any instructions to produce a short textual summary.  
       OUTPUT: A concise textual description/metadata.

3. (Conceptual Bundle/Pack Step)  
   – Because there is no direct “zip file creation” skill, the subagent must conceptually “package” the snippet files and the text summary into one final location or artifact. For example, it might save them on a server in one folder or produce a “final assets” link containing all snippet files and the summary.  
   – If the pipeline environment supports custom packaging, it can be done behind the scenes here, returning a single link or reference to that final deliverable.

4. (Subagent Final Output)  
   – Return the final “packaged” reference (e.g., a single location or a conceptual “zip file” link) plus the textual summary.

E) SILOS  
• SILO 1: Receive Snippets and Metadata  
  – Input the 3 snippet URLs and track metadata.  
• SILO 2: Generate or Merge Summary  
  – Use #223 (LLM) if we want a dynamic text summary.  
• SILO 3: Bundle into Final Output  
  – Conceptually gather everything; produce a single final artifact link or “zip” concept; return that plus the summary.

This completes the subagent’s flow for final packaging and distribution of the three 16-beat snippet audio files.